Lane Departure Detection Based On Multi-vision

In the Advanced Driver Assistant Systems of vehicle,lane detection is an essential part of it.At present,vision-based lane detection methods mostly use single front-view cameras to acquire road images,and they can meet the requirements of lane detection in the situation without congestion in general,but they can't get good results in congested conditions with too many vehicles and people.Therefore,it is of great significance to study a lane detection method that is suitable for all kinds of road conditions for improving the overall detection accuracy of actual road environment.In order to solve the problems mentioned above of the traditional lane detection methods,a new method was presented which applies a front-view camera and two top-view cameras to collect road images and detect lanes simultaneously.In this method,images gained by two top-view cameras were used to provide supplementary lane information for the front-view images,so that the mistake rate and miss rate of lane detection can be reduced.The main work of this paper is as follows:Image preprocessing methods were used for the collected road images.Traditional methods of region of interest division were improved to improve the accuracy of the division.Gray processing was used and the images enhancement processing was performed on the top-view road images.Various types of filtering methods were analyzed and compared in detail.Finally,4×4 median filter was used.The OSTU threshold segmentation and the Canny edge detection were performed.The lanes in the front-view images were fitted through the B-spline curves based on the RANSAC algorithm,and the lanes in the top-view images were detected by the improved Hough algorithm.The confidence level of the lane information detected by the front-view camera were judged,and then the top-view lane information was supplied for the front-view images according the judgment to achieve multi-camera information fusion and the purpose of driver assistance.Lane acquisition and detection experiments collected a variety of driving video and got a large number of road images through the conversion.Lane detection method based on multi-vision was verified by experiments and the results showed that it is practical and able to meet the requirements of lane detection under complex road environment.